Research On Knowledge Base Question Answering System Based On Deep Learning

The goal of the knowledge base question answering(KBQA)system is to make the machine understand the user's questions and then use the knowledge base to give logical and accurate answers.The existence of large-scale knowledge base provides a convenient and quick basis for finding answers to questions.In recent years,the application of deep learning theory has been extended to different fields,which also promoted the rapid development of the knowledge base question answering system field.At present,the knowledge base question answering system has more research on the quality of the answer generation,but lacks the attention to the context information generated by the answer.There are problems such as insufficient information utilization,weak expression of emotional features,and lack of model fusion mechanism.This paper has carried out research work from the following aspects,and has achieved certain research results:In this paper,the deep learning framework is used to realize the learning algorithms of neural network,convolutional neural network and recurrent neural network models.This paper also achieves a simple question and answer task of the knowledge base.In this paper,the evaluation index parameters of the experimental test data are given and the advantages and disadvantages of different models are summarized and analyzed.It is pointed out that the generalization ability of the neural network model is poor.Convolutional neural networks and circulating neural networks are prone to overfitting.Recurrent neural networks contain global and sequential information,which is more suitable for dialogue scenarios with emotional features.This paper proposes a question-and-answer dialogue generation model based on deep learning.The model is built in the framework of seq2 seq model.By establishing the reward function mechanism and using the gradient strategy algorithm to optimize,the model's answer generation effect is improved.The dialogue generation model reduces the probability of the universal answer during the conversation.At the same time,the fluency of the dialogue is improved,which is conducive to the continuous conversation of the dialogue.This paper proposes a sentiment classification and generation model based on deep learning.The model uses deep learning techniques to extract the emotional characteristics of the text,and adds this emotional feature as a supervised signal to the answer generation process.The model learns the emotional transfer distribution from the conversational conversation and adds this sentiment distribution feature as a supervised signal to the decoder,causing the dialog system to generate an answer with emotional characteristics.In this paper,the above models are tested on various data sets,and the experimental test results are in line with the expected assumptions,which proves the validity of the proposed model.